A boutique data
science practice for
teams who need the complex
parts done simply.
Seed & Series A founders, growth-stage product teams, and Fortune 500 operators who want difficult work executed properly.
Jonathan Dinu. Pearson author, founder of Zipfian Academy (acq. Galvanize), research @ CMU, UC Berkeley CS & Physics.
Services.
4–24 weeks
Generative AI & ML systems.
Custom diffusion and transformer pipelines, RAG, boutique agents, fine-tuned vision models — trained, evaluated, and deployed.
Bayesian modeling you can defend.
Hierarchical models, probabilistic forecasting, and uncertainty quantification — delivered with the error bars your board will ask about.
Causal inference, done properly.
Interupted time series, diff‑in‑diff, probabilistic latent variable models — for teams that need to know what actually moved the number.
Data strategy & architecture.
Scalable data pipelines, distributed ML, and a 90-day roadmap your CFO will fund. Informed by a decade of shipping systems on Spark, Ray, and everything in between.
Fractional Chief Data Officer.
Hiring, roadmap, tooling, board narrative. Drawing on years leading academic programs at Galvanize and Zipfian. Typically six months to a full‑time hand-off.
Live training & workshops.
Two-day intensives for product, engineering, and exec teams. Drawn from years of developing custom in-person programs at Zipfian Academy and online courses for Pearson's Live Training platform.